[Snowball-discuss] Minor(?) errors in snowball's generated C code

[Tom Lane]

Olly Betts <olly at survex.com> writes:
> On Sun, Sep 30, 2018 at 12:55:40PM -0400, Tom Lane wrote:
>> It's not clear to me if this is actually a bug, or it's more of a case of
>> "we don't care whether this happens in this code path".  Coverity does have
>> a lot of heuristic checks that sometimes give false positives.

> It's not a bug - it gives signal f, which will propagate upwards until
> either something handles it (and anything which does will restore c to
> the saved value at that point), or we return to the (non-snowball) caller
> (in which case the position of c isn't important).

OK.

> I guess we could always restore z->c here since the value will get reset
> later when it matters anyway, and this is not a code path that's taken
> in normal operation.  But this seems a somewhat dubious check and I'm
> hesitant to start changing the code generation for such things.

No, I'd not suggest changing things if you think this is fine as-is.
As I said, it's only a heuristic check ...

			regards, tom lane